🌟 Dataset Playground — Machine Learning Datasets for Practice (CSV)

Welcome to the repository of ready-to-use CSV datasets for machine learning practice, data cleaning, feature engineering, and exploratory data analysis (EDA). This collection includes tabular, classification, regression, text (sentiment), and time-series datasets designed for both learning and experimentation — from classics such as Iris and Titanic to messy customer, sales, review, and sensor data for realistic cleaning exercises.

✨ What’s inside?

This repo contains 9 structured datasets across common ML scenarios:

  • Customer behavior and spend prediction
  • Fraud detection
  • Sentiment analysis
  • Time-series forecasting
  • Classic ML benchmark datasets
  • Real-world business and academic examples

📏 Size at a Glance

File Rows Columns
customer_data.csv 1,000 9
sales_data.csv 1,000 8
reviews_data.csv 1,000 6
sensor_data.csv 1,000 5
College.csv 777 19 (the first, unnamed column holds the college name)
Iris.csv 150 5
Purchase.csv 10 4
Salary.csv 205 2
Titanic.csv 891 12

Row counts exclude the header row.

📊 Dataset Catalog

File Theme Problem Type Target/Goal Best For
customer_data.csv Customer analytics Regression Predict Purchase_Amount EDA, data cleaning, feature engineering
sales_data.csv Transactions Classification Detect Is_Fraudulent Fraud detection, anomaly handling
reviews_data.csv Product feedback Multiclass classification Predict review sentiment NLP, text preprocessing
sensor_data.csv IoT sensor readings Regression Predict Temperature / Humidity Time-series modeling, missing data handling
College.csv Higher education Classification Predict college outcomes / categories (for example Private) Feature selection, logistic regression
Iris.csv Botanical dataset Classification Identify flower species Beginner ML, clustering, classification
Purchase.csv Customer purchase behavior Classification Predict purchase decision Marketing analytics
Salary.csv Employment dataset Regression Predict salary from experience Linear regression, simple modeling
Titanic.csv Passenger data Classification Predict survival Data wrangling, binary classification

🧩 Detailed Dataset Overview

1) customer_data.csv — Customer Spending Data

Focus: Missing values, duplicates, inconsistent data, categorical encoding, text cleaning
Target: Purchase_Amount
Problem type: Regression
Goal: Predict how much a customer is likely to spend based on demographics and behavior.

Suggested features:

  • Age
  • Gender
  • Country
  • Signup_Date
  • Email / Phone pattern-derived features

All columns: Customer_ID, Name, Age, Gender, Email, Phone, Country, Signup_Date, Purchase_Amount. Values are intentionally messy: Gender mixes M, F, Male, Female, and blanks; 519 of the 1,000 Age values and 172 Country values are empty.

2) sales_data.csv — Sales Transactions

Focus: Outlier detection, numerical normalization, date/time handling, large-number transformations
Target: Is_Fraudulent
Problem type: Classification
Goal: Identify whether a transaction is fraudulent.

Suggested features:

  • Transaction_Amount
  • Payment_Method
  • Discount_Applied
  • Product_Category
  • Transaction_Date

3) reviews_data.csv — Product Reviews

Focus: Text preprocessing, sentiment analysis, categorical variables
Target: Sentiment
Problem type: Multiclass classification
Goal: Classify reviews into positive, neutral, or negative sentiment.

Suggested features:

  • Review_Text
  • Rating
  • Review_Date
  • Customer_ID

4) sensor_data.csv — Sensor Readings

Focus: Time-series missing values, outliers, scaling
Target: Temperature or Humidity
Problem type: Regression
Goal: Forecast sensor behavior using historical event patterns.

Suggested features:

  • Timestamp
  • Humidity
  • Pressure
  • Derived temporal features like hour, day, and month

5) College.csv — College Data

Focus: Institutional metrics, academic performance analysis
Problem type: Classification / exploratory analysis
Goal: Understand relationships between admission data and college outcomes.

Typical variables include:

  • Private, Apps, Accept, Enroll
  • Top10perc, Top25perc, Outstate
  • Room.Board, Books, Expend, Grad.Rate

Other columns: F.Undergrad, P.Undergrad, Personal, PhD, Terminal, S.F.Ratio, perc.alumni. The first column has no header and contains the college name — load it with index_col=0 in pandas.

6) Iris.csv — Iris Flower Dataset

Focus: Classic classification learning
Target: species
Problem type: Classification
Goal: Predict the flower species using petal and sepal measurements.

Features:

  • sepal_length
  • sepal_width
  • petal_length
  • petal_width

7) Purchase.csv — Purchase Decision Dataset

Focus: Customer decision modeling
Target: Purchased
Problem type: Classification
Goal: Predict whether a customer purchases based on their profile.

Columns:

  • Employee
  • Age
  • Salary
  • Purchased (target)

This file has only 10 rows, so it suits quick demonstrations rather than model evaluation.

8) Salary.csv — Experience vs Salary

Focus: Simple supervised learning
Target: salary
Problem type: Regression
Goal: Estimate salary based on years of experience.

Features:

  • years_experience
  • salary

9) Titanic.csv — Titanic Survival Dataset

Focus: Data cleaning, feature engineering, survival analysis
Target: Survived
Problem type: Classification
Goal: Predict passenger survival probability using profile and travel information.

Key variables include:

  • Pclass, Sex, Age, Fare
  • SibSp, Parch, Embarked
  • Cabin, Ticket

Also included: PassengerId and Name.

🚀 Suggested Use Cases

This repository is ideal for:

  • Learning Python or R for data science
  • Practicing pandas, numpy, and scikit-learn
  • Building regression and classification pipelines
  • Cleaning messy datasets and handling missing values
  • Creating EDA notebooks and portfolio projects
  • Exploring feature engineering ideas

🐍 Quick Start with pandas

Use Python with pandas installed in your active environment:

python -m pip install pandas

Load a dataset directly from GitHub without cloning the repository:

import pandas as pd

base = "https://raw.githubusercontent.com/gcfernando/datasets/main/"

titanic = pd.read_csv(base + "Titanic.csv")
college = pd.read_csv(base + "College.csv", index_col=0)  # first column = college name

print(titanic.shape)          # (891, 12)
print(titanic.isna().sum())   # missing values per column

Or clone the repository and read the files locally:

git clone https://github.com/gcfernando/datasets.git
cd datasets

From that folder, load the local CSV files:

import pandas as pd

titanic = pd.read_csv("Titanic.csv")
college = pd.read_csv("College.csv", index_col=0)
print(titanic.shape)  # (891, 12)
  1. Start with EDA
  2. Clean missing values and inconsistent entries
  3. Encode categorical variables
  4. Engineer useful features
  5. Train baseline models
  6. Compare results and iterate

✅ Summary

This dataset collection is a compact, practical toolbox for machine learning and data analysis learning. It combines beginner-friendly datasets with more realistic business and time-series challenges, making it a great resource for students, researchers, and aspiring data scientists.

❓ FAQ

Are the customer, sales, review, and sensor datasets real data? They appear to be synthetic: for example, customer IDs are random UUIDs and the email domains look generated. Treat them as practice data, not as real people or real business results.

Which dataset should a beginner start with? Iris.csv for classification and Salary.csv for linear regression are small and clean. Move on to Titanic.csv for missing-value handling, then customer_data.csv and sales_data.csv for messier, more realistic cleaning.

Where did the classic datasets come from? Iris.csv, Titanic.csv, and College.csv are versions of widely used public teaching datasets. No source or license information is recorded in this repository, so check the original sources before redistributing them.

🔗 Related Projects

Read all Machine Learning Practice Datasets guides online · Project overview · All developer projects